Abstract: Existing event log sampling methods have the problem of low sampling efficiency when processing large-scale event logs collected by information systems.Therefore,a novel efficiently log sampling approach called as LogRank++ was proposed.The important characteristics of the trace were first confirmed,such as activity and direct follow relation.Then,the importance value of the calculated trace was sorted,and a group of the most important trace was selected to form a sample log.In addition,the efficiency of sampling technology was evaluated from two aspects: sampling quality and sampling efficiency.The proposed sampling approach had been implemented in the open-source process mining toolkit ProM.Experimental evaluation with both synthetic and real-life event logs demonstrated that the proposed sampling approach provided an effective solution to improve event log sampling efficiency as well as ensuring high quality of the obtained sample logs from a process discovery perspective.

Key words: LogRank++, log sampling, process discovery, quality measurement

摘要： 针对已有采样方法处理大规模事件日志时存在采样效率低的问题,提出一种高效的业务过程事件日志采样方法LogRank++。首先确定轨迹的重要性特征,然后对计算轨迹的重要性值进行排序,最后选择一组最重要的轨迹组成样本日志。综合采样质量和采样效率两方面来评估此采样方法的高效性。所提采样方法已在开源过程挖掘工具平台ProM中实现。实验分析表明,相比已有采样方法,在保证样本日志质量的前提下,LogRank++能够大幅提高日志采样效率。

关键词: 日志排序, 日志采样, 过程发现, 质量评估